Someone correct me if they disagree, but I think it is generally a bad
practice to try to infer things about time in a simulation using real time.
I always set up ecological simulations so that time is represented in steps,
such as one minute or one second per step. Using real time would assume that
the computer is doing nothing else other than running my simulation -- which
would require maximizing the thread priority. I could tie the steps into
real time (wait until the next second for the next step to occur or
something similar), or use an external MIDI timing device, but that would
only slow up the simulation and I would be using steps as time anyway. 

This is also why I am not a big fan of using separate threads or processors
for agents without synchronizing to a single scheduler. Not synchronizing
this way implies that the statistical distribution of delays among thread
processing or machines somehow represents that of delays in the real world.
If I want randomness, it is better to put it into the response to steps.   

I take comfort in that a physicist recently proposed that real time is
actually just a sequence of finite events...  

Cheers!

John